Decide
Which model, for what, at what cost
Start from the job rather than the rate card. Each task below works out what one run actually costs on every model that could do it — then the full catalogue is underneath, for when you already know which model you mean.
A vendor filter is on.Some rows and models are hidden. Use the filter in the header to show everything.
Start from the task
- Build a website“I want a website and I do not write code”from $0.1860 a run
- Change code that already exists“I need to change an app without breaking it”from $0.1300 a run
- Write and reply to email“I want help drafting and replying to email”from $0.00022 a run
- Make a presentation“I need slides from notes or a document”from $0.00600 a run
- Work on a spreadsheet“I want help with formulas, cleaning or analysing a sheet”from $0.00352 a run
- Summarise long documents“I have something long and I need the short version”from $0.00340 a run
- Research a topic“I need to understand something I know nothing about”from $0.0180 a run
- Review code“I want a second pair of eyes before I ship”from $0.00160 a run
- Write a LinkedIn post“I want to write something for LinkedIn that does not sound like AI”from $0.00102 a run
- Pull structured data out of documents“I need the numbers out of a pile of invoices or forms”from $0.00107 a run
- Work out why something is broken“Something is failing and I do not know why”from $0.00490 a run
- Automate something repetitive“I do this by hand every week and I want it to stop”from $0.0345 a run
The whole catalogue
64 models across 7 vendors, 50 of them with both figures read verbatim off the vendor’s own page. Prices areper million tokens (How model prices are quoted: dollars per million tokens, charged separately for what you send and what comes back.), in USD, as published — and a figure a vendor does not publish says so rather than showing a zero.
Showing all 64 models, cheapest input price first.
| Vendor | Cache read | Cache write 1h | Evidence | ||||
|---|---|---|---|---|---|---|---|
| GPT-5 Nano | OpenAI | $0.05 | $0.4 | $0.005 | not published | not published | Vendor page |
| Claude Haiku 5.5 | Anthropic | $0.1 | $0.5 | $0.01 | $0.2 | 1M | Vendor page |
| GPT-6 Luna | OpenAI | $0.1 | $0.5 | $0.01 | not published | not published | Vendor page |
| Gemini 2.5 Flash Lite | $0.1 | $0.4 | $0.01 | not published | not published | Vendor page | |
| GPT-4.1 nano | OpenAI | $0.1 | $0.4 | $0.025 | not published | 1.0M | Corroborated |
| GPT-4o mini | OpenAI | $0.15 | $0.6 | $0.075 | not published | 128K | Corroborated |
| GPT-5.6 Luna | OpenAI | $0.2 | $1.2 | $0.02 | not published | not published | Vendor page |
| GPT-5.4 Nano | OpenAI | $0.2 | $1.25 | $0.02 | not published | not published | Vendor page |
| Jamba Mini | AI21 | $0.2 | $0.4 | not published | not published | 256K | Vendor page |
| GPT-5 Mini | OpenAI | $0.25 | $2 | $0.025 | not published | not published | Vendor page |
| Gemini 3.1 Flash Lite | $0.25 | $1.5 | $0.025 | not published | not published | Vendor page | |
| Gemini 3.5 Flash Lite | $0.3 | $2.5 | $0.03 | not published | not published | Vendor page | |
| Gemini 2.5 Flash | $0.3 | $2.5 | $0.03 | not published | not published | Vendor page | |
| Codestral 25.08 | Mistral AI | $0.3 | $0.9 | $0.03 | not published | 256K | Corroborated |
| GPT-4.1 mini | OpenAI | $0.4 | $1.6 | $0.1 | not published | 1.0M | Corroborated |
| Gemini 3 Flash Preview | $0.5 | $3 | $0.05 | not published | 1.0M | Vendor page | |
| GPT-3.5-turbo | OpenAI | $0.5 | $1.5 | not published | not published | 16.4K | Corroborated |
| GPT-5.4 Mini | OpenAI | $0.75 | $4.5 | $0.075 | not published | not published | Vendor page |
| Gemini 3.8 Flash | $0.75 | $3.75 | $0.075 | not published | not published | Vendor page | |
| Gemini 3.6 Flash | $0.75 | $3.75 | $0.075 | not published | not published | Vendor page | |
| Gemini 3.1 Flash Live Preview | $0.75 | $4.5 | not published | not published | 131.1K | Vendor page | |
| Claude Haiku 4.5 | Anthropic | $1 | $5 | $0.1 | $2 | not published | Vendor page |
| GPT-5.1 | OpenAI | $1.25 | $10 | $0.125 | not published | not published | Vendor page |
| GPT-5 | OpenAI | $1.25 | $10 | $0.125 | not published | not published | Vendor page |
| Gemini 2.5 Pro | $1.25 | $10 | $0.125 | not published | not published | Vendor page | |
| Grok 4.20 (Non-Reasoning) | xAI | $1.25 | $2.5 | $0.2 | not published | 1M | Corroborated |
| Gemini Omni Flash Preview | $1.5 | $17.5 | not published | not published | 131.1K | Vendor page | |
| GPT-5.2 | OpenAI | $1.75 | $14 | $0.175 | not published | not published | Vendor page |
| GPT-5.3 Codex | OpenAI | $1.75 | $14 | $0.175 | not published | 400K | Corroborated |
| Claude Sonnet 5.5 | Anthropic | $2 | $10 | $0.1 | $4 | 1M | Vendor page |
| Claude Sonnet 5 | Anthropic | $2 | $10 | $0.2 | $4 | not published | Vendor page |
| GPT-6.1 Sol | OpenAI | $2 | $10 | $0.1 | not published | not published | Vendor page |
| GPT-6 Sol | OpenAI | $2 | $10 | $0.2 | not published | not published | Vendor page |
| GPT-5.6 Terra | OpenAI | $2 | $12 | $0.2 | not published | not published | Vendor page |
| Gemini 3.1 Pro Preview | $2 | $12 | $0.2 | not published | not published | Vendor page | |
| Jamba Large | AI21 | $2 | $8 | not published | not published | 256K | Vendor page |
| GPT-4.1 | OpenAI | $2 | $8 | $0.5 | not published | 1.0M | Corroborated |
| GPT-5.4 | OpenAI | $2.5 | $15 | $0.25 | not published | not published | Vendor page |
| Command R+ | Cohere | $2.5 | $10 | not published | not published | 128K | Vendor page |
| GPT-4o | OpenAI | $2.5 | $10 | $1.25 | not published | 128K | Vendor page |
| Claude Sonnet 4.6 | Anthropic | $3 | $15 | $0.3 | $6 | not published | Vendor page |
| Claude Sonnet 4.5 (latest) | Anthropic | $3 | $15 | $0.3 | not published | 200K | Corroborated |
| Claude Opus 5.5 | Anthropic | $4 | $20 | $0.2 | $8 | 1M | Vendor page |
| GPT-5.6 Sol | OpenAI | $4 | $20 | $0.4 | not published | not published | Vendor page |
| GPT-5.6 | OpenAI | $4 | $20 | $0.4 | not published | 1.1M | Vendor page |
| Claude Opus 5 | Anthropic | $5 | $25 | $0.5 | $10 | not published | Vendor page |
| Claude Opus 4.8 | Anthropic | $5 | $25 | $0.5 | $10 | not published | Vendor page |
| Claude Opus 4.7 | Anthropic | $5 | $25 | $0.5 | $10 | not published | Vendor page |
| Claude Opus 4.6 | Anthropic | $5 | $25 | $0.5 | $10 | not published | Vendor page |
| Claude Opus 4.5 | Anthropic | $5 | $25 | $0.5 | $10 | not published | Vendor page |
| GPT-5.5 | OpenAI | $5 | $30 | $0.5 | not published | not published | Vendor page |
| gpt-rosalind-discovery | OpenAI | $5 | $25 | $0.5 | not published | not published | Vendor page |
| gpt-rosalind-research | OpenAI | $5 | $25 | $0.5 | not published | not published | Vendor page |
| Claude Fable 5.1 | Anthropic | $10 | $50 | $0.25 | $20 | 1M | Vendor page |
| Claude Fable 5 | Anthropic | $10 | $50 | $1 | $20 | not published | Vendor page |
| Claude Mythos 5.1 | Anthropic | $10 | $50 | $0.25 | $20 | not published | Vendor page |
| GPT-6 Astra | OpenAI | $10 | $50 | $1 | not published | not published | Vendor page |
| GPT-4 Turbo | OpenAI | $10 | $30 | not published | not published | 128K | Corroborated |
| gpt-5.6-cyber | OpenAI | $12.5 | $75 | $1.25 | not published | 400K | Vendor page |
| GPT-5 Pro | OpenAI | $15 | $120 | not published | not published | 400K | Corroborated |
| GPT-5.2 Pro | OpenAI | $21 | $168 | not published | not published | 400K | Corroborated |
| GPT-5.5 Pro | OpenAI | $30 | $180 | not published | not published | not published | Vendor page |
| GPT-4 | OpenAI | $30 | $60 | not published | not published | 8.2K | Corroborated |
| GPT-5.4 Pro | OpenAI | $30 | $180 | not published | not published | 1.1M | Corroborated |
By vendor
What the evidence column means
- Vendor page
- Both figures appeared verbatim on the vendor’s own page, in this model’s own row.
- Corroborated
- Two independent price databases agree. The vendor’s own page did not confirm it.
- One source
- One price database lists it and nothing else corroborates it.
- Sources disagree
- Sources give different figures. Both are shown and neither was chosen.
Only the first two appear in this table. Anything a single database asserts, or that sources disagree about, is published separately atunconfirmed models.50 of 64 models publish no 1-hour cache-write price.How we verify a price.